Pedestrian Detection and Distance Estimation Using Thermal Camera in Night Time

Abstract

In this paper, we propose a method to detect a pedestrian in real time in a low illumination environment and estimate the distance from the camera using a smart phone based thermal camera. Thermal cameras use equipment that can be attached to low-cost smartphones and which use cameras for image processing in real-time. A pedestrian detector is created using a multi-stage cascade learning device to detect pedestrians in a low-illuminated environment, and the pedestrian area is detected using this detector. Then, the distance is estimated by calculating the position of the pedestrian detected in the real-world 3D environment in the 2D thermal image by calculating the parameters detected by the thermal imaging camera in advance. Experimental results show that the detection accuracy of pedestrians is about 91% and the accuracy of distance estimation is 95%. In this way, the proposed method can be applied to the image sensing system in real time in a low-illuminance environment such as nighttime.